Análise comparativa de métodos de aprendizado profundo para predição de preços de contratos futuros de milho

Abstract

Deep learning is a modern machine learning approach widely used to solve complex and evolving problems. This work presents a comparative analysis of different deep learning models applied to the prediction of price variations in corn futures contracts. The methodology involves testing multiple architectures for each model, selecting the best configurations, and subsequently comparing them to a baseline model built on a simple forecasting heuristic. The main contribution of this study is to deepen the understanding of the applicability of deep learning models in financial time series forecasting, with potential extensions to other domains. The results were satisfactory, with convolutional and recurrent models consistently outperforming the baseline, while the LSTM (Long-Short Term Memory Neural Network) model delivered lower-than-expected performance, highlighting the impact of architectural complexity in scenarios with more limited datasets.